DeepSeek has consolidated its fast, expert, and image recognition modes into a single unified intelligent system. The model now automatically detects query complexity and activates visual capabilities when images are provided, adjusting processing power based on the specific task requirements.
The company plans to release the V4.1 Flash model around September 10, 2026. According to internal and external testing, this new version outperforms the V4 Pro across key metrics, including performance, cost-efficiency, processing speed, and total execution time.
Following the launch of V4.1 Flash and prior to the release of V4.1 Pro, DeepSeek will route all V4 Pro requests to the V4.1 Flash model. During this period, users will be billed according to the pricing structure of the V4.1 Flash model.
